Lqas PDF
Lqas PDF
Lqas PDF
Field Manual
27 April 2012
Table of contents
Executive Summary (Quick Guide to LQAS) ...................................................................................................... 4 This manual: what it is and who it is for4 What is LQAS? ............................................................................................................................................... 4 Lot definition ................................................................................................................................................. 4 Sampling methodology ................................................................................................................................. 4 Cluster selection ....................................................................................................................................... 4 House selection......................................................................................................................................... 4 Child selection........................................................................................................................................... 5 Questionnaire administration................................................................................................................... 5 Interpreting the results ................................................................................................................................. 5 Ensuring quality of the survey ....................................................................................................................... 6 Communicating results and taking action..................................................................................................... 6 Limitations of LQAS ....................................................................................................................................... 6 Chapter I LQAS within GPEI ........................................................................................................................... 8 Use of LQAS within the Global Polio Eradication Initiative (GPEI) ................................................................ 8 Importance of immunization quality assurance ........................................................................................... 8 What will LQAS provide and not provide? .................................................................................................... 8 LQAS, Independent Monitoring (IM) and EPI cluster surveys ....................................................................... 8 Clustered-LQAS............................................................................................................................................ 10 Chapter II Implementation of a clustered-LQAS urvey ................................................................................ 13 Overview ..................................................................................................................................................... 13 Phase I Planning the survey...................................................................................................................... 13 Objective ................................................................................................................................................. 13 Lot definition........................................................................................................................................... 13 Cluster definition .................................................................................................................................... 15 Selection of lots ...................................................................................................................................... 15 Timing of the survey ............................................................................................................................... 15 Target population ................................................................................................................................... 15 Ascertaining vaccination status (definition of failure) ........................................................................ 15 Recommended interpretation framework: ....................................................................................... 16 Additional thresholds.............................................................................................................................. 17 Ascertaining awareness about the campaign ......................................................................................... 17 Phase II - Conducting the survey ................................................................................................................. 19 Sampling clusters in the lot (First stage)................................................................................................. 19 Sampling households in the cluster (Second stage) ............................................................................... 19 Sampling one child per household (Third stage) .................................................................................... 22 Data collection ........................................................................................................................................ 23 Taking Action .......................................................................................................................................... 23 What if... ................................................................................................................................................. 24 Summary of sampling procedure (six clusters of 10) ............................................................................. 25 Communicating results and taking action .............................................................................................. 25
Chapter III - Composition of the survey teams and responsibilities ............................................................... 26 Coordinators................................................................................................................................................ 26 Supervisors .................................................................................................................................................. 26 Surveyors ..................................................................................................................................................... 26 Local Guides ................................................................................................................................................ 27 Chapter IV Training ....................................................................................................................................... 28 Clustered-LQAS field survey manual ........................................................................................................... 28 Training workshop....................................................................................................................................... 28 Checklist before going to the field .............................................................................................................. 28 Supervisor/coordinators check-list: Best Practices to ensure quality during the stages of clustered-LQAS . 30 ANNEXES .......................................................................................................................................................... 31 ANNEX I Tools for data collection............................................................................................................. 31 Example of questionnaire ..................................................................................................................... 31 Example of LQ summary sheet ............................................................................................................... 32 Table of random numbers ..................................................................................................................... 33 ANNEX II - Budget and logistic arrangements ............................................................................................. 35 ANNEX III PPS and other methods of sampling ........................................................................................ 36 Sampling with probability proportionate to the size (PPS) .................................................................... 36 Other methods........................................................................................................................................ 37 References................................................................................................................................................... 39
What is LQAS?
Lot definition
Lot should be defined, based on geographical or administrative boundaries, usually at district or sub-district level, where there is a responsible officer present and accountable for corrective actions.
Sampling methodology
In each lot, 60 children will be sampled in six different clusters, each of 10 children (e.g., villages, settlements).
Cluster selection
In each lot, six clusters should be selected, based on PPS (Probability Proportionate to the Size) methodology (i.e., if a village has a larger population, it is more likely to be selected).
House selection
In each cluster, 10 houses with eligible children should be selected, with the following methodology: 1) Divide the locality in four sectors using an available map, or sketch one identifying some landmarks (e.g. road, river, school, mosque, church, etc.). 2) Select one sector randomly and go to the selected sector. 3) If the sector has maximum 20 households (HHs), number them and select one randomly as the starting point of the survey. 4) If the sector has more than 20 HHs, repeat steps 1-3 until a sector with maximum 20 HHs is obtained. 5) Administer the survey in the HH selected as the starting point. 6) Once the survey is completed in the selected HH, turn right exiting the house and
Locality
Sector 1 Sector 2
River
Road
2 1 6 12 16 19 13 7 8 14 17 20
3 9
4 10 15
5 11
18
Sector 3
Sector 4
select the following HHs to survey according to the predetermined sampling interval. 7) Complete the survey of 10 individuals in the selected cluster and move to the following cluster.
20 households
Leave 2 households in the areas of high density with more than 20 households (e.g. Urban areas): S 1 S 2 3 4 S 5 6 7 S S S S S 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
> 20 households
Child selection
Only one child should be selected in each house: 1. Ask the head of the household or a caregiver how many eligible individuals are present in the house. 2. List them and select one randomly to administer the questionnaire. 3. Once the questionnaire is administered move to the following household according to the sampling interval.
Questionnaire administration
1. Fill out questions regarding the cluster location when you arrive in the locality. 2. The most important question is if the individual selected is vaccinated (the child has the finger mark). 3. You may want to add other questions to the questionnaire depending on the survey needs. 4. Always allow for a comments section in the questionnaire to specify if there were any issues with selecting households this will be important for evaluating the ease of the survey procedure for future applications.
Interpretation Framework
Sample = 60 (6 clusters of 10)
Nr Unvaccinated
89
60
If too many lots are failed, the programme can use additional thresholds (d=19) to identify lots with the lowest coverage (i.e. lack of evidence that it is above 60%) and others (i.e. above 60%). The key to obtaining reliable results is the quality of survey. The supervisor should train surveyors and monitor the activity, especially to ensure (1) that the cluster is selected based on the PPS; (2) the households are selected randomly based on the methodology; and (3) only one child is selected in each house.
Limitations of LQAS
When the outcome of LQAS is discussed and used to make programmatic decisions, the following limitations should be noted. 1. There is some risk of misclassification Due to the relatively small sample size and clustering approach, there is a chance of misclassification. In fact in lots where coverage varies greatly between the six clusters, the statistical errors (i.e. alpha and beta error) can be very high (up to 20%). The use of multiple bands for classification may further increase the error of misclassification of lots.
2. LQAS may not reflect the coverage of entire lots Especially, if the lot is too large and heterogenic in coverage, the outcome of LQAS may not be applicable to the entire lot (e.g., districts). 3. LQAS does not provide an exact coverage estimate LQAS provide a classification of SIA quality, but it does not give us the point estimate of coverage. 4. LQAS is a household-based survey LQAS reflects the coverage level of children who are at home at the time of the survey.
Finally the EPI cluster survey methodology, based on a much larger cluster-sample, should be used if users want to estimate vaccination coverage with a 95% confidence interval rather than having the classification obtained with LQAS 1. Table I.2 Comparison of LQAS, IM and EPI cluster sampling Methodology Objectives Advantages Lot quality assurance To provide information -Small sample sizes are sampling. on whether vaccination generally used. coverage is adequate or -Useful for monitoring inadequate. performance. - Fairly rapid to conduct. - Provides actionoriented information. -Results are interpretable statistically when sample is drawn randomly. Disadvantages - Follows a strict probabilistic methodology that is complicated to implement. - Less familiar for users in the field. - Demands more resources than IM devoted to sampling in the field. - Does not determine actual coverage at lot level.
Independent Monitoring.
To provide information on multiple indicators that can be used to guide improvements to reach more children by enabling corrective action both during SIAs and in planning for the next rounds.
- Based on targeted sampling. - Rapid, inexpensive, easy to implement. - More familiar for users in the field. - Provides rapid feedback for local decision-making on multiple indicators in problem areas (coverage, reasons for non-vaccination, social indicators).
- Results cannot be interpreted statistically. -Does not provide a statistically reliable estimation of coverage (e.g. with a 95% confidence interval), but rather a quick validity check. - Cannot be generalized outside of the area where implemented. - Less useful in areas where coverage is consistently low (i.e. it tends to over-state coverage). - Provides an estimate of - Follows a strict vaccination coverage probabilistic with 95% confidence methodology based on interval. cluster sampling that is more complicated to - Useful for postimplement. campaign evaluation purposes. -Sample size is much - Results are larger than LQAS
Users interested in IM or EPI cluster sampling are referred to the specific guidelines/manuals: - GLOBAL GUIDELINES on Independent Monitoring of Polio Supplementary Immunization Activities (SIA) available from the Global Polio Eradication Initiative - Immunization Coverage Cluster Survey Reference Manual available from the World Health Organization
Clustered-LQAS
Traditionally, LQAS has been proposed with a simple random sample design. This means that, if the sample size required to perform the assessment is N=60, then 60 individuals need to be sampled completely randomly in the lot. However, if a lot is large geographical area sampling 60 individuals completely randomly would require considerable time and resources and make the assessment unpractical. To increase the practicality and rapidity of the assessment, Clustered Lot Quality Assurance Sampling (Clustered-LQAS) has been developed which divides the sample (N) into smaller clusters (k) of n individuals each. For example, if N=60, k=6, n=10, we would need to select six villages randomly in the health district, and in each village select ten individuals randomly, instead of conducting up to 60 trips in 60 different locations. Figure I.1 Classic LQAS and Clustered-LQAS
Logically, this would increase the rapidity of the assessment. However, there is a decrease in precision, especially if coverage between the clusters is very heterogeneous. NB: The clustered-LQAS classification is relevant to the complete lot sample (N) and not to the cluster (n). In other words, if a lot is rejected, corrective actions will have to be taken for the entire lot (e.g. the district) and not the location from which the cluster was drawn (e.g. the village in a district).
Table I.3 Advantages and disadvantages of the 6x10 Clustered-LQAS compared to classic LQAS Advantages Increased rapidity and feasibility in the field. Possibility to assess larger lots because less travel is involved (only six locations). Disadvantages Loss of precision if coverage in the lot is not homogeneous (i.e. there is a great variation of coverage between the six clusters in the lot). It is not possible to classify the clusters in terms of vaccination coverage, but the classification can only apply to the whole lot.
LGA
= LOT
Ward1
Ward2
Ward3
Ward4
Ward5
Ward6
House 1 Child 1
House 2 Child 2
House 3 Child 3
House 4 Child 4
House 5 Child 5 1
House 6 Child 6
House 7 Child 7
House 8 Child 8
House 9 Child 9
House 10 Child 10
= ONE CLUSTER
Figure II.2 Example of Clustered-LQAS sampling frame proposed in Pakistan
District
UC
= LOT
Settlement 1
Settlement 2
Settlement 3
Settlement 4
Settlement 5
Settlement 6
House 1 Child 1
House 2 Child 2
House 3 Child 3
House 4 Child 4
House 5 Child 5 1
House 6 Child 6
House 7 Child 7
House 8 Child 8
House 9 Child 9
House 10 Child 10
= ONE CLUSTER
What to consider when defining the lots:
1) Size of the lots (i.e. consider at what level you want/can implement control measures) a) If the lots are large (e.g. LGAs with a total population larger than 100 000), it may not be feasible to implement control measures. Furthermore, the bigger the lot is, the higher the chance will be of high heterogeneity of vaccination coverage in the area (see point 2 below). b) If the lots are small (e.g. the area covered by one health centre with a total population of between 3 000 - 5 000), the control measures may be too localized to have an impact on the vaccination programme. In addition, significant resources to cover all areas of interest would be required. 2) Homogeneity in terms of vaccination coverage (i.e. select a territory where the likelihood of being vaccinated is similar across the different areas) 2 a) Consider choosing areas covered by one vaccination team/supervisor as part of the same lot. b) Consider separating densely-populated areas and scarcely-populated areas into different lots. c) Consider separating rural and urban areas into different lots. Cluster definition In the clustered-LQAS approach described in this manual, the sample for each lot is divided into smaller clusters (k) from which a predetermined number of individuals (n) will be sampled. Depending on the definition of a lot, the cluster could be obtained from the lowest administrative division present that allows individuals to be selected. Normally the clusters can be defined as the localities with clear administrative boundaries present in the lot. For example, clusters can be the villages or catchment areas in a health district. In order to select the clusters, you will need to have a list of all these administrative localities in your lot, ideally with census information. Selection of lots One way to select the areas at risk is to perform a risk assessment using the information available (e.g. reports from previous campaigns). In the GPEI context, the areas generally selected as lots are: Polio-infected areas (e.g. areas with reported polio cases in the previous six months) Other areas defined as high-risk by the local programme. Timing of the survey In the GPEI context, we recommend that clustered-LQAS surveys be conducted a few days (less than one week) after the completion of the specific round to be assessed, and that the survey be completed within a day or two. Target population In the context of evaluating polio SIAs, the target population will be the same one targeted by the SIA: Children below five years of age (0-59 months) living in the area defined as the lot during the campaign. NB: in case the vaccination campaign targets different age groups you may want to consider surveying this age group also with the LQAS survey. Ascertaining vaccination status (definition of failure) During polio SIAs an indelible mark is put on the finger-nail of every child vaccinated with OPV.
This is because in the clustered-LQAS approach we are selecting clusters in the lot assuming that they will be representative of the entire lot. In other words we do not expect the coverage in the clusters to vary greatly from the mean coverage in the lot (ideally the Standard Deviation should not exceed 0.1).
Any child who does not have the finger mark should be considered as not vaccinated. Verbal history should not be considered Recommended interpretation framework The approach recommended by the GPEI to assess the quality of OPV coverage is based on a band approach that allows us to classify lots in three bands (High, Medium, and Low coverage) using two decision values (d) in a sample (N) of 60 divided into six clusters of 10 children. The GPEI acknowledges that coverage levels of at least 90% are probably necessary to stop polio transmission, but also recognizes that having only one threshold (90%) with one decision value does not provide information on how far the lot is from this coverage target. Hence, the recommended interpretation framework allows us to classify the lots, loosely based on two levels of coverage (90% and 80%) and on two consequent decision values (3 and 8) using a sample of 60 (6x10). Decision rule and corrective actions are as follows: 1. 0-3 unvaccinated (out of 60): coverage is in the Higher Band (i.e. some evidence that coverage may be reaching 90%) PASS a) Decision: data do not suggest performance gaps. b) Recommended action: No action needed. 2. 4-8 unvaccinated (out of 60): coverage is probably in the Medium Band (i.e. probably between 80 and 90%) WARNING a) Decision: data suggest performance gaps, though these may not be highest priority if there are areas of greater concern. b) Recommended action: review other indicators (e.g. in-process monitoring or administrative coverage) in order to decide how to increase coverage levels above 90% in the lot. 3. Nine or more unvaccinated (out of 60): coverage is probably in the Lower Band (i.e. we lack strong evidence that it is above 80%) FAIL a) Decision: data suggest serious performance gaps these areas should be given highest priority in the short term. b) Recommended action: consider re-doing the SIA (entire lot or selected areas, based on other indicators).
Interpretation Framework
Sample = 60 (6 clusters of 10)
Nr Unvaccinated
89
60
Additional thresholds The GPEI recognizes that in some areas vaccination coverage is consistently very low (most lots would fail in the classification above). For programmatic reasons, an additional threshold (19 out of 60) may be applied to identify areas with highest priority (i.e. 20 or more children are missed in the sample of 60). This means we lack strong evidence that it is above 60%. Ascertaining awareness about the campaign The approach presented in this manual is focussed mainly on assessing the quality of vaccination coverage, although LQAS can be used with any dichotomous variable. In parallel to vaccination activities, intensified communication and social mobilization activities take place in high-risk areas prior to the immunization campaign. LQAS could then also be used to assess the quality of communication activities using the same framework. In order to do so the awareness variable should be coded as a dichotomous variable (i.e. as a yes/no question such as were you aware of the vaccination campaign?). In this sense, any caregiver interviewed who indicates that she or he was not aware of the campaign prior to the arrival of the vaccination teams should be considered as a failure (i.e. not aware). In this case we will use only one of the decision rules, classifying the lots as follows: Decision rule and corrective actions: 1. 0-3 not aware (out of 60): some evidence that awareness may be reaching 90% PASS a) Decision: data do not suggest performance gaps. b) Recommended action: No action needed. 2. 4 or more not aware (out of 60): we can not be confident that awareness is above 90% FAIL
a) Decision: data suggest performance gaps in the communication campaign, though these may not be highest priority if there are areas of greater concern. a) Recommended action: consider a different communication strategy.
We describe the sampling procedure based on the fixed sample size of 60 children aged from 0 to 59 months divided in six clusters of 10 children.
sectors of approximately equal population size. They will select one sector randomly. They will go to the selected sector (figure IV.1). Sectors of 20 households or fewer: If the number of households in the sector is 20 or fewer, surveyors will number them and randomly select one to be the starting point of the survey. Sectors of more than 20 households: If there are more than 20 households in the sector, the surveyors will divide the sector into four smaller sectors of equal size and randomly select one, according to the procedure described above, and repeat this procedure until they obtain a sector with 20 or fewer households. If this is not feasible (i.e. a map is not available or the field conditions prevent one being sketched on site), the surveyors will go to the centre of the sector and use the spin-the-bottle procedure. Figure II.4 Selection of the first household to survey in a village/neighbourhood
Locality
Sector 1 Sector 2
River
Road
2 1 6 12 16 19 13 7 8 14 17 20
3 9
4 10 15
5 11
18
Sector 3
Sector 4
Example: The surveyors divided the locality in four sectors of approximately 20 or fewer households using the available landmarks (a river and a road); they randomly selected Sector 4; they numbered all the households in that sector and randomly selected Household 13 as the starting point of the survey.
Figure II.5 Example of a sketch drawn to select the first household in a village/neighbourhood with more than 20 houses
Selection of the nine following households in the survey To ensure that the sample is spread across the entire locality, households immediately next to each other in the cluster should not be selected. Once surveyors have visited the first household, they should continue with the survey by exiting the door to their right and selecting the nine subsequent households, leaving a predetermined interval between them (Figure IV.3). This interval depends on the size of the settlement: 20 households: turn to the right and leave one household between each selected household >20 households: turn to the right and leave two households between each selected household Figure III.6 Procedure to select the following households in a village/neighbourhood
20 households
Leave 2 households in the areas of high density with more than 20 households (e.g. Urban areas): S 1 S 2 3 4 S 5 6 7 S S S S S 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
> 20 households
If, while following this procedure, surveyors end up in a different sector of the map from the one where they have selected the first household, it is not a problem as long as they remain in the same locality (village or neighbourhood) chosen for the cluster. If they have walked the entire locality without interviewing all the participants required for the cluster, they then should go to the next nearest locality (village or neighbourhood) in the same lot to complete the survey for that cluster. Sampling one child per household (Third stage) Only one child between 0 and 59 months should be selected randomly per household to respond to the survey. To ensure that the participant in the house is selected randomly, all children present in the house eligible for the vaccine should be numbered and one selected using a random ballot (see Table of Random Numbers in Annex I).
The surveyors will ask an adult to answer the questions on his/her behalf. If no adults are present in the house, they will move on to the neighbouring one to the right. If in the selected household there are no eligible individuals they will also go to the neighbouring house to the right. Data collection The LQAS questionnaires list the questions to be asked in each household and provide space to record information about the individuals interviewed. The main variable of interest for this survey will be the presence/absence of the finger mark on the nail of the child selected (see Questionnaire in Annex I). In addition to paper questionnaires, where available, mobile phones can also be used for data collection and transmission of results to ensure real-time analysis and avoid intentional and unintentional errors 4. Summary sheets may also be useful for analyzing and reporting data, where the number of vaccinated and unvaccinated individuals is recorded for each cluster in the lot. Examples of LQ questionnaires and Summary Forms are available in ANNEX I Tools for data collection. Taking action Once the survey is completed and 60 individuals are sampled action can now be taken according to the framework (see Phase II Choosing the clustered-LQAS framework) that has been chosen for use.
Summary sheets Questionnaires
WHO has developed a technology (Episurveyor) that allows conducting a Clustered - LQAS survey using mobile phones (see separate WHO manual Setting up a Clustered - LQAS Survey Using Mobile Phones, George Walker, August 30 2010)
What if... In following this procedure you may encounter conditions that vary slightly from what is planned. This is not a problem as long as you decide beforehand how to treat unusual situations in a standard way. Examples of unusual situations: a) The building selected is a religious or commercial building, an office, shop, or school. Do not administer the survey. Go to the neighbouring household in the agreed direction (e.g. to your right), without leaving an interval. b) There are buildings with more than one household (e.g. apartment blocks in urban centres). You will have to select one household randomly in the building following a stepwise approach. Select one floor randomly, then on the selected floor select one apartment randomly. To avoid clustering all the individuals in one building, once a household is selected, leave the building and return to the street, and follow the sampling interval to the right. c) The household selected is empty. Go to the next neighbouring household in the agreed direction without leaving an interval. d) There are only children in the household. Go to the next neighbouring household in the agreed direction without leaving the interval. e) There is nothing to your right when you exit the house. In this case turn to your right until you face a direction where you see households. f) In turning to your right and following the procedure you reach a wall or a physical barrier. You will have to turn to your right on site until you face a direction where you see households. g) In turning to your right and following the procedure you end up covering the same area that you have already covered. You will have to ignore this area and proceed to the closest area that you have not yet covered. If all the locality has been covered, consider moving to the closest locality in the same lot (see point i) below) h) You reach the dead end of the road (i.e. there are no other roads to your side). You will have to turn approximately 180 to the right and proceed on the other side of the road (e.g. the left side of the road if you were on the right side). i) In following the procedure you are not able to complete the interviews because there are not enough households or target individuals in the locality. Go to the geographically closest locality (village or neighbourhood) in the same lot to continue the interviews and complete the cluster. Select the first household of the locality and the subsequent ones according to the same procedure described. j) You find children who are not permanent residents of the house (e.g. visitors). They should be included if they were eligible to receive the vaccine (aged between 0-59 months living in an area targeted by the SIA).
Summary of sampling procedure (six clusters of 10) 1) First Stage a) Six clusters per lot will be pre-selected by the survey coordination team using PPS. b) Select one locality randomly in each of the six cluster areas. c) Go to the selected locality. 2) Second Stage a) Divide the locality in four sectors using an available map, or sketch one identifying some landmarks (e.g. road, river, school, mosque, church, etc.). b) Select one sector randomly and go to the selected sector. c) If the sector has maximum 20 households (HHs), number them and select one randomly as the starting point of the survey. d) If the sector has more than 20 HHs, repeat steps a-c. e) Administer the survey in the HH selected as the starting point. f) Once the survey is completed in the selected HH, turn right exiting the house and select the following HHs to survey according to the predetermined interval. g) Complete the survey of 10 children in the selected cluster and move to the following one. If completing an LQAS for more than one vaccine, complete the survey with 10 eligible individuals for each vaccine. 3) Third Stage: Selecting one child per household a) In each household one child should be randomly selected to respond to the survey. b) List and number all the children from 0-59 years of age living in the household. c) Select one child randomly from the list using the table of random numbers. d) The survey in the lot is completed once 60 children per lot are surveyed (six clusters; 10 children per cluster).
Communicating results and taking action Especially if we are in the FAIL or WARNING bands, it is very important to forward this information immediately to decision-makers and interested stakeholders so that measures to raise vaccination coverage can be rapidly implemented at lot level. You will need to consider what may have caused the low coverage identified in the rejected lots (e.g. absence of immunization cards rather than vaccine, not enough vaccine vials in that area, presence of hard-to reach areas, etc.) and then design solutions accordingly. What if the lot PASSES? Clustered-LQAS may also show that vaccination coverage is higher than expected in certain lots, meaning that the classification is: PASS. If the lots thought to have low coverage are actually doing well, users may want to consider redirecting resources to areas more in need. This decision needs to be taken carefully however, making sure that the workforce is not removed from areas that still need it.
Coordinators
The professionals involved in protocol writing and planning the survey (e.g. local WHO staff, international WHO staff, or consultants) can act as survey coordinators making sure that the teams can cover all the territory. The LQAS coordinator is the person in charge of the survey (from recruiting the human resources, coordinating the training and the implementation, interpreting the results). Coordinators are also responsible for ensuring overall quality of the survey (i.e. quality in training, in data collection, in data entry, etc.). The coordinator is also responsible for writing the final report of the survey.
Supervisors
At state/province level, supervisors should be nominated to ensure the quality of LQAS. The supervisor officer should support introduction, monitor the quality of LQAS and corrective actions. Two Supervision scenarios are normally considered for LQAS surveys: 1) The survey is supervised by local WHO staff. 2) The survey is supervised by other independent partners, such as NGOs, sub-contractors, universities, etc. Where possible: Employ experienced staff from WHO, UNICEF and international partners as supervisors. Outsource the task to external organizations only when similar quality of staff can be provided. Staff involved in vaccination activities (i.e. local ministry of health staff) should not be involved in the coordination of the survey.
Surveyors
Depending on the number of lots, their geographical size, the timing of the evaluation and conditions in the field, the coordinators will determine the number of surveyors needed for each lot. A team of two surveyors may be able to cover a lot (e.g. completing six clusters) in two days if provided with a vehicle but this may not be the case depending on resources available and on the country. Responsibilities of surveyors: 1) 2) 3) 4) Visiting the household and administering the questionnaire. Should be knowledgeable about the territory that they need to cover. Should be very familiar with the customs and culture of the people living in the area. Should speak the local language.
Surveyors work efficiently in teams of two, ideally a man and a woman. In some areas, where men are not allowed in the households, only female surveyors in your workforce may be preferred.
To ensure independence of surveyors from vaccination activities they should: 1) Be deployed to locations outside their daily responsibility. 2) Regularly change the areas that they assess (e.g., different provinces after different rounds).
Local guides
In some cases, local guides may be recruited to support the surveyors with data collection in territories with which there are unfamiliar.
Chapter IV Training
Clustered-LQAS field survey manual
A survey manual has to be prepared so that the survey teams know what to do once they are in the field. Based on what was discussed above on the sampling plan and on the sampling frame, the manual will include instructions on: 1. Introduction to the study and the objectives. 2. How to choose sampling points. 3. How to choose one target individual in each household. 4. How to administer the questionnaire. 5. Who to contact while in the field. 6. To whom to communicate the decision in order for actions to be taken.
Training workshop
Before starting data collection, the surveyors will need a training session (workshop) of one or two days on data collection, reporting the findings, and decision-making if necessary. The training should combine theoretical and practical activities. You will need to present the information contained in the manual to the survey teams during this workshop to make sure that everybody understands the purpose and objectives of the survey. The training workshop should include the following activities: 1. An overview of the survey, its purpose, how it will be carried out, and how the data will be used. 2. Review of instructions, with examples of hypothetical problems that data collectors might encounter. 3. Review of data collection forms, their purpose and content, and practice in completing them with hypothetical information. 4. A practice session in the field (select an area near to where the training is taking place) so that surveyors understand how to select random starting points and individuals in the cluster. This session should last up to half a day, leaving time to discuss any problems encountered in the field in order to adapt the survey manual. During the field practice exercise, each surveyor should do a few samplings themselves 5. Written examination (short quiz) and group discussion of the results to make sure that everybody has understood what to do in the field If the conduction of the survey is outsourced to an external organization, WHO/UNICEF staff should still make sure to attend to ensure quality.
It is important to ensure that everybody will travel to the lots well-prepared after the workshop. You need to make sure that survey teams bring with them to the field the following tools: 1. Personal identification they should always be able to demonstrate who they are and why they are there to the interviewees. 2. Survey manual for reference in the field. 3. List of the localities (clusters) in the lot to survey if it is not possible at central level to pre-select the clusters where the survey will take place, then the surveying teams will need to select the six clusters once they are in the lot with the help of local public health officials. 4. Maps of the localities to survey: if maps of the localities (clusters) of the lots are available in advance, it may be helpful to hand them to the surveying teams during training; if they are not available they will have to draft them on site.
5. Table of random numbers: to allow random selection of sectors in a village (or settlement) with more than 20 households, individuals, etc. 6. Notepad and pen: to take notes, draft the map of the locality if needed, complete the questionnaires, etc. 7. Questionnaires: carry some spare copies in case the questionnaires get lost or damaged during travelling. 8. Summary sheet: also carry some spare copies in case the questionnaires get lost or damaged during travelling. 9. Survey contact sheet: to know who to contact during the survey, especially if a decision is taken in their lot.
Supervisor/coordinators check-list: Best practices to ensure quality during the stages of Clustered-LQAS
Stage Best practice
Employ experienced staff from WHO, UNICEF and partners. Outsource the task only when the organization can provide the similar quality of staff. A surveyor should be deployed to locations outside his/her responsibility. A surveyor should regularly change geography (e.g., different provinces). The training of surveyors should include both theory and field practice. During field practice, each surveyor should do a few samplings personally. If outsourced, WHO/UNICEF staff should attend the training to assure the quality of training and surveyor. Lots and clusters (e.g., settlements) should be pre-selected. Surveyors should follow the pre-determined house selection methods (e.g. random starting points, skip one/two houses, one child per house). Use mobile phones to transmit results to ensure real-time analysis and avoid intentional and unintentional errors. The dedicated officer should support introduction, monitor the quality of LQAS and corrective actions (e.g. "refresher course"). Supervisors should be nominated at state/province level to ensure the quality of LQAS.
Methodology
Quality assurance
ANNEXES
ANNEX I Tools for data collection
Example of questionnaire
Q1: Lot: _________________________ Q2: Cluster Number: |__| (1-6) Q3: Name of the locality for the cluster (village/neighbourhood): ________________________________ Q4: Name of Surveyor: ________________________________ Signature: ____________ Q5: Name of LQAS Supervisor: __________________________ Signature: _____________ Q6: Date: |__|__| |__|__| |__|__| Vaccination Status Note: Only administer the questionnaire to one eligible child randomly selected per household. Progressive number of household in the 1 2 3 4 5 6 7 8 9 10 cluster Q7: Age In months; if above 59 months do not administer questionnaire. Q8: Sex M=Male F=Female Q9: Vaccinated (Presenting finger Y=Yes/ N=No mark) Q10: Was caregiver aware of the polio vaccination campaign Y=Yes/ N=No prior to the arrival of the vaccination teams? Q11:Comments_____________ _____________ _____________ ___________ ________________________ ____________ _________________________________________
/10
/60
Supervisors Name ____________________________ Signature __________________ Supervisors Comments: _______________________________________________________________________________________ _______________________________________________________________________________________ _______________________________________________________________________________________ _______________________________________________________________________________________ ___________________________
How to use the table of random numbers The table of random numbers is a list of randomly-selected numbers. To select a number randomly, you need to know how many households are included in your village/sector and how many digits this total number of households has. For example, if the surveyor needs to select randomly a household in a village, he or she first needs to number all households in the sector between 1 and n, where n is the total number of households in the sector. For example, let us say that the surveyors are in a sector with 20 households and have to select one randomly as the starting point of the survey. They number all the households from 1 to 20. Twenty is a two digit number so they will have to choose randomly from the table the first number with two digits between 1 and 20. They point with the pen to a random spot on the table (one way to do it is to place a finger on the table without looking at it before). From this spot, move your finger towards the right until you find the first 2-digit number between 01 and 20. The household allocated this number will be the one selected to start the survey. Look at the table. Let us say that the number 8 from the cell 87131 (row 7 column 7) has been pointed to. Move to the right and look at every 2-digit number in the table. This produces the following numbers: 87, 71, 13, 31, 14, 42... Hence, 13 is the first 2-digit number between 1 and 20 screened. Consequently, the randomly selected household for this sector will be number 13.
Example of PPS sampling In this example for a health sub-district with a population of 5841, we plan to select 5 villages in each of which 10 individuals will be interviewed. The surveyors compiled the sheet below. Sheet to select the villages in the health district according to PPS Number of clusters needed: 5 Cumulative population of the health district: 5841 Sampling interval: 5841/5= 1168 Random number (between 1 and the sampling interval): 661 1. Village No. 2. Village Name 3. Village Population 4. Cumulative Population 5. No of clusters selected in the village 6. Progressive Sampling Number
1 A 246 246 2 B 1577 1823 661* 3 C 468 2291 1829** 4 D 340 2631 5 E 220 2851 6 F 246 3097 2997*** 7 G 190 3287 8 H 1124 4411 4165**** 9 I 61 4472 10 J 154 4626 11 K 139 4765 12 L 60 4825 13 M 14 4839 14 N 38 4877 15 O 19 4896 16 P 41 4937 17 Q 120 5057 18 R 455 5512 5333 19 S 51 5563 20 T 26 5589 21 U 199 5788 22 V 21 5809 23 W 32 5841 *Random number; **progressive number 1829 (random number 661 + sampling interval 1168); *** progressive number 2997 (previous progressive number 1829 + sampling interval 1168); **** etc. According to this procedure the five villages selected to sample 10 individuals will be: B, C, F, H, and R.
Other Methods
In case PPS is not viable (e.g. you do not have a list of localities with population data) you can consider other methods to sample clusters in the lot. Simple random sampling (SRS) For SRS you just need to have the list of all localities in the lot and you will select the number needed by simple random ballot. The advantage of SRS is that it does not require census information. SRS will give the
same chance to every locality to be selected, although the chance of each child being selected will depend on the locality. This results in a sample that is less representative of the entire lot. Geographical sampling methods If maps of the area considered as a lot are available, you may opt for geographical sampling methods, such as centric systematic area sampling (CSAS), to ensure that the probability of being selected as cluster is evenly spread across the whole territory. The CSAS method involves dividing the survey area (lot) into a grid of non-overlapping numbered squares of equal size (quadrants), selecting randomly the quadrants from which we will select one cluster each, and sampling the community located closest to the centre of each quadrant [4]. NB: We discourage the use of these methods because they produce a sample that is not representative of the entire lot. They should only be used when sampling with PPS is not possible.
References
(1) Dietz V, Venczel L, Izurieta H et al. Assessing and monitoring vaccination coverage levels: lessons from the Americas. Revista Panamericana de Salud Pblica 2004 December;16(6):432-42. (2) World Health Organization. Immunization coverage cluster survey - reference manual. Geneva: World Health Organization, 2005. (3) Robertson SE, Valadez JJ. Global review of health care surveys using lot quality assurance sampling (LQAS), 1984-2004. Social Science & Medicine 2006 September;63(6):1648-60. (4) Myatt M et al. A field trial of a survey method for estimating the coverage of selective feeding programmes. Bulletin of the World Health Organization 2005 January;83(1):20-6. (5) Lemeshow S, Taber S. Lot quality assurance sampling: single- and double-sampling plans. World Health Statistics Quarterly 1991;44(3):115-32. (6) Pezzoli L, Andrews N, Ronveaux O. Clustered lot quality assurance sampling to assess immunisation coverage: increasing rapidity and maintaining precision. Tropical Medicine & International Health 2010 May;15(5):540-6. (7) Sandiford P. Lot quality assurance sampling for monitoring immunization programmes: costefficient or quick and dirty? Health Policy & Planning 1993 September;8(3):217-23. (8) Staff of the Computation Laboratory. Tables of the cumulative binomial probability distribution. Cambridge: Harvard University Press, 1954. (9) Pezzoli L et al. Cluster-sample surveys and lot quality assurance sampling to evaluate yellow fever immunization coverage following a national campaign, Bolivia, 2007. Tropical Medicine & International Health 2009 March;14(3):355-61.